Academy of Management Journal 2005, Vol. 48, No. 2, 346–357. EXISTING KNOWLEDGE, KNOWLEDGE CREATION CAPABILITY, AND THE RATE OF NEW PRODUCT INTRODUCTION IN HIGH-TECHNOLOGY FIRMS KEN G. SMITH University of Maryland, College Park CHRISTOPHER J. COLLINS Cornell University KEVIN D. CLARK Villanova University A field study of top management teams and knowledge workers from 72 technology firms demonstrated that the rate of new product and service introduction was a function of organization members’ ability to combine and exchange knowledge. We tested the following as bases of that ability: the existing knowledge of employees (their education levels and functional heterogeneity), knowledge from member ego networks (number of direct contacts and strength of ties), and organizational climates for risk taking and teamwork. The introduction of new products and services is There is an important symbiosis between these a critical determinant of organizational perfor- two knowledge streams. For example, it takes some mance and survival (Damanpour, 1991). By intro- level of existing knowledge or know-how to de- ducing new products and services, organizations velop new knowledge, and this new knowledge can establish new markets and technologies must at some point lead to new products or services (Burgelman, 1991) and adapt and change to meet to provide value (Boland & Tenkasi, 1995; Harga- new market demands (Brown & Eisenhardt, 1995). don & Fanelli, 2002). Thus, existing knowledge in- A key premise in the literature on new product fluences the extents to which new knowledge is innovation is that the rate of new product introduc- created, and the new knowledge that is formed is tion is a function of a firm’s ability to manage, converted to existing knowledge in the form of new maintain, and create knowledge (Cohen & products and services. The present research exam- Levinthal, 1990; Drazin & Rao, 2002). ined the relationship between existing knowledge Hargadon and Fanelli (2002) divided the re- in an organization and both the organization’s search on knowledge into two streams or ap- knowledge creation capability and how this capa- proaches. First, there are those studies that center bility influences the introduction of new products on how knowledge is distributed among a firm’s and services. We pursued two research questions: employees, technologies, resources, routines, and (1) How does the existing or accessible knowledge procedures. The emphasis of this effort has been of a firm impact the knowledge creation capability somewhat static, focusing on how existing knowl- of the firm? (2) With existing knowledge taken into edge can be replicated and exploited to affect cer- account, how does the firm’s knowledge creation tain outcomes, such as financial performance (Lev- capability affect its level of innovation? To investi- itt & March, 1988). The second stream has been gate existing or available knowledge in a firm, we more dynamic, emphasizing how knowledge, and focused on the stock of knowledge held by the especially new knowledge, leads to the generation members of its top management team (TMT) and of novel organizational outcomes, such as new key knowledge workers and the distribution of products (Kogut & Zander, 1992). knowledge among these firm members; the infor- mation and knowledge that are accessible to the We thank Ming-Jer Chen, Ed Locke, Patrick Maggitti, TMT and knowledge workers’ through their “ego Mike Pfarrer, Cindy Stevens, Wenpin Tsai, three anony- networks”; and the guidance provided through the mous reviewers, and Donald Bergh for helpful comments organization’s “climate.” Examining knowledge on drafts of our manuscript and Dean Howard Frank and creation capability, we emphasized the firm mem- the Robert H. Smith School of Business for financial bers’ ability to combine and exchange information support for this study. to obtain new knowledge (Nahapiet & Ghoshal, 346 2005 Smith, Collins, and Clark 347 1998). We studied innovation by examining levels 248). When individuals who hold different levels of new products and services firms introduced to and kinds of knowledge begin to combine ideas, the market in 72 high-technology companies. TMT they create new potential knowledge. When this members included key officers and executives who new potential knowledge is validated, for example were part of their CEOs’ decision-making teams, by test marketing or other experiments, it is con- and knowledge workers included employees who verted into new knowledge. were critical to creating new knowledge or devel- Our review of the knowledge literature suggests oping innovations within their organizations (Bo- at least three categories of organizational resources land & Tenkasi, 1995). Our unit of analysis in this that impact knowledge creation capability. First are research was the organization. stocks of individual knowledge in an organization, which Hargadon and Fanelli (2002) referred to as latent knowledge. Second are ego networks, or re- THEORY AND HYPOTHESES lational contacts, which facilitate knowledge flows The concept of organizational knowledge is between employees and stakeholders by creating fuzzy and has been defined in a number of ways (cf. access and motivation to exchange ideas and infor- Huber, 1991; Nonaka & Takeuchi, 1995). Following mation (Hargadon & Fanelli, 2002; Nahapiet & Nonaka and Takeuchi, we defined organizational Ghoshal, 1998). Finally, there are the organiza- knowledge as the validated understanding and be- tional routines and processes that comprise a firm’s liefs in a firm about the relationship between the climate that informally, and perhaps tacitly, define firm and its environment. In this definition, orga- how the firm is to develop and use knowledge. We nizational knowledge is static, reflecting current theorize that a firm’s stocks of TMT and knowledge viewpoints on how existing resources should be worker knowledge, ego networks, and organiza- configured and exploited for advantage. Further, tional climate affect knowledge creation capability, we assumed that organizational knowledge is com- which in turn will affect the creation of new products posed of two types: explicit knowledge, defined as and services. codified and easily translated facts and informa- tion; and tacit knowledge, defined as personal Knowledge Stocks and Knowledge Creation know-how that may be hard to confirm and convey Capability (Nonaka & Takeuchi, 1995; Polanyi, 1975). While this static view of knowledge is important, Most studies of organizational learning recognize researchers have also taken a more dynamic per- employees as a primary repository of organiza- spective on knowledge, emphasizing that the cre- tional knowledge (Argote, 1999). Indeed, the natu- ation of new knowledge is essential for the success ral abilities, intelligence, and skills of key employ- and survival of firms competing in dynamic envi- ees acquired from formal education and job ronments (Kogut & Zander, 1992; Nonaka & Takeu- experience constitute the level of an organization’s chi, 1995). This literature suggests that organiza- human capital (Becker, 1964). For the purposes of tional knowledge creation is dependent on the this research, we defined a stock of organizational ability of organization members to exchange and knowledge as the years of industry experience and combine existing information, knowledge, and education of a firm’s TMT members and knowledge ideas (Kogut & Zander, 1992, 1996). Following Na- workers and as the diversity of the information and hapiet and Ghoshal (1998), we defined and mea- knowledge this group holds (as reflected in their sured an organization’s knowledge creation capa- functional backgrounds). We considered experi- bility as the extent to which TMTs and knowledge ence, education, and knowledge diversity a “stock” workers have access to one another and other stake- because they represent the amount of knowledge in a holders, are capable of combining information and firm at a certain point in time (Dierickx & Cool, 1989). knowledge into new knowledge, and perceive Experience. New product knowledge resides in value from the exchange and combination process. the minds of the managers and knowledge workers Implicit in the notion of exchange is the assump- responsible for such innovations as new products tion that individuals hold different levels and types (Drazin & Rao, 2002). The knowledge these workers of knowledge and information, and that they can/ hold is often tacit and noncodifiable (Glaser, 1984), will engage in teamwork and communication to developing and expanding as they spend more time learn from one another even when payoffs are un- in specific jobs and industries. Therefore, we argue certain. Combination refers to the process of bring- that organizations with TMTs and knowledge ing together “elements previously unconnected or workers who have extensive work experience in an by developing new ways of combining elements industry will have greater expertise and thus more previously associated” (Nahapiet & Ghoshal, 1998: relevant knowledge to bring to the exchange and 348 Academy of Management Journal April combination process. Further, researchers have dis- Finally, employees with greater levels of education tinguished between the knowledge and the knowl- are likely to be more receptive to new ideas and edge-processing capabilities of experts and nov- change (Boeker, 1997). Thus, we predict: ices. Experts have larger knowledge bases, developed through their experiences in specific job Hypothesis 1b. The number of years of educa- domains, a better understanding of how to apply tion of a firm’s TMT and knowledge workers is their knowledge, and knowledge structures that are positively associated with the firm’s knowledge larger and more accessible than those of novices creation capability. (Glaser, 1984; Lord & Maher, 1990). Bearing out Functional heterogeneity. Although level of or- these notions, empirical research has shown that, ganizational knowledge, expressed in experience compared to novices, experts have richer and more and education, will likely impact what is brought detailed schemata to use in decision making, to the combination and exchange process, the vari- greater relevant knowledge to recall, an ability to ety of types and levels of knowledge organization focus more on inconsistencies in information, and members hold may also be important (Kimberly & less bias in their recall of information (Fiske, Evanisko, 1981). Essentially, when individuals Kinder, & Larter, 1983; McKeithan, Reitman, Ru- within a group hold different information, cogni- eter, Hirtle, 1981). tive conflict is likely to increase, which can lead to Similarly, Hambrick and Mason (1984) argued more productive exchanges and greater attempts to that key organization members carry their job-re- combine information and knowledge in an effort to lated experiences as part of their cognitive make- reduce conflict (Nemeth, 1992). Cohen and ups and can draw upon this experience in decision Levinthal (1990) argued that the greater the unique making. They noted that organizations run by ex- knowledge held by individuals in a firm, the ecutives with limited experience will have “re- greater the potential for new knowledge to be gen- stricted knowledge bases” upon which to draw. erated by knowledge exchange. Conversely, when Cohen and Levinthal (1990) specifically described all individuals in an organization hold the same how a lack of investment in individual knowledge stock of knowledge, creativity may be dampened and expertise could bar the development of new because members will be less likely to perceive knowledge. In view of these arguments, we predict: value in the exchange and combination process Hypothesis 1a. The number of years of experi- (Amabile, 1996). We hypothesize that greater diver- ence of a firm’s TMT and knowledge workers in sity in the stocks of knowledge held by a firm’s a firm is positively associated with the firm’s TMT and knowledge workers will yield greater va- knowledge creation capability. riety in the information brought to the exchange and combination process. This view is consistent Education. As with work experience, develop- with the TMT literature, which suggests that mental psychologists support the connection be- greater diversity in a TMT will lead to a wider tween education level and improved knowledge range of strategic options and greater creativity in structures and information processing. In particu- decision making (Boeker, 1997). lar, Glaser (1984) argued that changes in knowledge base through education could produce sophisti- Hypothesis 1c. The level of functional hetero- cated changes in cognitive performance. Education geneity of a firm’s TMT and knowledge work- helps individuals improve their understanding of ers is positively associated with the firm’s what they know, more accurately predict outcomes, knowledge creation capability. better manage time and resources, and monitor re- sults. In effect, education provides new explicit infor- Ego Networks and Knowledge Creation mation and knowledge that greatly influence an indi- Capability vidual’s cognitive reasoning skills. Hambrick and Mason (1984) similarly argued Knowledge creation often depends on the com- that workers’ formal educations mirror their knowl- munication within a firm’s community of experts edge bases and cognitive abilities. Bantel and Jack- (Boland & Tenkasi, 1995). It follows then that how son (1989) argued that better-educated TMTs key employees are connected to one another and to would have stronger cognitive abilities and as a important stakeholders in social relations or net- result generate more novel and creative organiza- works will be an important indicator of the knowl- tional outcomes. Similarly, the findings of Kim- edge they can draw upon in the exchange and com- berly and Evanisko (1981) suggested that greater bination process (Nahapiet & Ghoshal, 1998). education led to greater innovation by improving Hansen (2002) argued that network relations are cognitive processing and problem-solving ability. important to knowledge creation because they in- 2005 Smith, Collins, and Clark 349 form network members about the existence, loca- Hypothesis 2b. The range of TMTs’ and knowl- tion, and significance of knowledge contained in a edge workers’ contacts in a firm is positively network and provide an important conduit for the related to the firm’s knowledge creation capa- flow of knowledge. We studied the ego-centered bility. networks of TMT members and knowledge work- The strength of network ties. The strength of a ers. Each network consisted of a focal manager or tie refers to the nature of a relational contact knowledge worker and a set of “category alters” (Granovetter, 1973). Closeness, long duration, and connected to the focal person (Wasserman & Faust, 1 frequent contact are characteristics of strong ties. In 1994). Specifically, we examined the TMT mem- this research, we measured the strength of the tie bers’ and knowledge workers’: (1) numbers of direct for each relationship in the network. In general, contacts, (2) ranges of different contacts, and (3) egos (TMT members and knowledge workers) will strength of ties. We examined the contacts of each be more likely to trust those alters with whom they TMT member and knowledge worker in relation to a have strong ties. They will be more likely to share predefined and bounded set of stakeholders. Our fo- knowledge and information with the latter than cus was on the relationship of each ego (a top man- with those with whom they have weaker ties, ager or knowledge worker) to a set of alters that were where trust is less evident. Although weak ties may potential sources of knowledge and information from provide certain efficiency benefits, especially when outside their TMT or set of knowledge workers. the meaning of information is not problematic Number of direct contacts. One of the most com- (Granovetter, 1973) or when networks are used for mon measures of an individual’s set of social rela- search activities (Hansen, 1999), strong ties are crit- tions is the number of people to whom he or she is ical when information is important, uncertain, or directly connected (Burt, 1982). Knowledge bene- ambiguous. Significant evidence suggests that fits from having a large number of direct contacts when ties are strong, individuals will be more will- include unique information, more information, and ing to exchange information and cooperate for mu- faster information (Burt, 1992). Thus, compared to tual benefit (Krackhardt, 1992). TMT members and knowledge workers with few direct contacts, those with many direct contacts Hypothesis 2c. The strength of ties in the will be able to obtain information faster, access richer TMTs’ and knowledge workers’ sets of rela- sets of data, and draw from broader sets of referrals, tions in a firm is positively related to a firm’s all of which should facilitate knowledge combination knowledge creation capability. and exchange (Nahapiet & Ghoshal, 1998). Hypothesis 2a. The number of TMTs’ and Organizational Climate and Knowledge Creation knowledge workers’ direct contacts in a firm is Capability positively related to the firm’s knowledge cre- The embedded knowledge and procedural infor- ation capability. mation captured in an organization’s climate is im- portant because they serve as a strategic expression Network range. Network range refers to the to the firm’s employees and stakeholders of how scope of different types of contacts contained in things are to be done and prioritized (Schneider, TMT and knowledge worker networks (Wasserman 2000). Organizational climate is defined as the col- & Faust, 1994). The narrower the range of contacts, lective attitudes and beliefs of employees about the the more limited the types of information and manner in which they perform their daily jobs knowledge one can draw upon. Networks compris- (Ashkanasy, Wilderom, & Peterson, 2000). Climate ing broader ranges of contacts, however, will have in this sense is an organizational attitude, reflecting more heterogeneous bases of information and embedded strategic values, beliefs, and assump- knowledge to draw on (Burt 1982). Access to more tions about how the organization should function diverse knowledge may enhance the possibility of (Schneider, 2000).2 Drawing from O’Reilly, Chat- combining and exchanging new information and man, and Caldwell (1991), we examined two as- may also increase the likelihood that an organiza- tion will gain value from this process. 2 Debate in the literature about the relationship be- tween climate and culture is ongoing. We do not contrib- 1 Ego-centered networks are also referred to as “per- ute to this debate; instead, we take the position that the sonal networks.” These networks are relational but are concepts of climate and culture are similar and that they incomplete since connections from each ego are only represent overlapping explanations of the same phenom- measured to some alters (Wasserman & Faust, 1994). ena (see Ashkanasy et al., 2000). 350 Academy of Management Journal April pects of organizational climate: the extent to which Bridges, & O’Keefe, 1984).3 Previous research on organizations encourage risk taking versus control, innovation supports the connection between new and the extent to which organizations emphasize knowledge creation capability and development of team behaviors versus individual behaviors. new products and services. For example, Dough- Climate for risk taking. For exchange and com- erty, Munir, and Subramaniam (2002) argued that bination to occur, organization members must per- innovation is dependent upon creative solutions ceive the willingness of the organization to exper- and accumulation of new knowledge in an organi- iment with new ideas and to take risks in both their zation. Hargadon and Sutton (1997) suggested that development and implementation (Nahapiet & knowledge is imperfectly spread across individuals Ghoshal, 1998). A climate that favors risk taking in an organization and that ideas from one group will encourage employees to test and exchange un- can solve the problems of another if exchanges are usual knowledge and ideas. Weick and Westley made between the groups. They further noted that when these exchanges are made, existing ideas from (1996) proposed that a climate emphasizing rules one group appear new to the other, and vice versa, and controls would push an organization toward resulting in potentially new products or services. order and away from learning and new knowledge Nonaka and Takeuchi (1995) were more precise creation. In contrast, a climate that stresses risk in detailing how the knowledge creation process taking and experimentation will move the organi- leads to new products and services. They argued zation toward disorder and experimentation that that through exchange and combination, tacit exist- leads to new knowledge creation. ing knowledge is transformed into explicit knowl- edge. Explicit knowledge takes the form of new Hypothesis 3a. A climate that stresses risk tak- “metaphors.” A metaphor is a way of understand- ing (as opposed to control) is positively related ing one image by thinking representatively of an- to a firm’s knowledge creation capability. other image (Nonaka & Takeuchi, 1995). Nonaka and Takeuchi described the following: “This cre- Climate for teamwork. Although a climate for ative, cognitive process continues as we think of the similarities among concepts and feel an imbal- risk taking may be important for new knowledge ance, inconsistency, or contradiction in their asso- creation, it is also important that norms of cooper- ciation, thus often leading to the discovery of new ation and teamwork exist in an organization for meaning” (1995: 67). Once new ideas and concepts exchange and combination to occur. Nahapiet and are explicitly developed through the discussion of Ghoshal (1998) argued that an atmosphere of coop- metaphors, these ideas are converted into actual eration opens access among group members and models and prototypes. The final step requires that creates individual motivation to exchange knowl- these new models or prototypes be tested and val- edge with group members. For example, Starbuck idated: hence the potential new knowledge is vali- (1992) described how norms for openness and dated and justified. teamwork in knowledge-intensive firms facilitated disclosure of information and loyalty building. Hypothesis 4. The knowledge creation capabil- Tushman and O’Reilly (1997) found that a climate ity of a firm is positively associated with the number of new products or services it introduces. of teamwork was key to effective creativity, and Amabile (1988) found that creativity was hurt In summary, we have argued that stocks of exist- when an organization’s climate was characterized ing organizational knowledge in TMTs and knowl- by a lack of cooperation. edge workers, information and knowledge from TMT and knowledge worker ego networks, and or- Hypothesis 3b. A climate that stresses team- ganizational climate will influence a firm’s knowl- work (as opposed to individualism) is posi- edge creation capability, and that knowledge cre- tively related to a firm’s knowledge creation ation capability will, in turn, affect the level of new capability. products and services. Accordingly, we expect that the knowledge creation capability is a necessary Knowledge Creation and Levels of New Product and Service Introduction 3 This definition is distinct from that for an organiza- tion’s new knowledge creation capability, which in- A firm’s level of innovation has often been de- volves the process of combination and exchange of infor- fined and measured as the number of new products mation among individuals to generate new firm or services it generates in a given period (Ettlie, knowledge. 2005 Smith, Collins, and Clark 351 requirement to the innovation process. In other R&D spending was not different from the R&D words, without the creation of new knowledge, spending of those that did respond (t169 1.03, there cannot be innovation. We therefore predict: n.s.). An interview with the CEO of each participating Hypothesis 5. The knowledge creation capabil- firm enabled us to gain her or his support for full ity of a firm fully mediates the relationship participation in the study, identify all the members between the firm’s existing knowledge, ego net- of the firm’s top management team and up to 15 works and climate, and number of new prod- knowledge workers, and collect information on the ucts and services. new products and services introduced in the last year as well as other background information on METHODS the company. The questionnaires that were distrib- uted to the TMT and knowledge workers were We examined firms’ knowledge creation capabil- identical in all respects, except that the surveys for ity with a field study of high-technology firms. Data knowledge workers included organizational cli- were collected from three key sources: (1) detailed mate items. An average of 3.52 TMT members (a 56 questionnaires completed by TMTs and knowledge percent internal response rate) and 5.95 knowledge workers, (2) a structured interview with the CEO of workers (a 58 percent internal response rate) re- each firm, and (3) archival data from company sponded from each firm. records. The CEO of a firm identified the TMT members who were part of the CEO’s decision- making team, and the knowledge workers included Variable Definition and Measurement those individuals that the CEO identified as being 4 Measures of stocks of knowledge (experience, ed-critical to knowledge creation and innovation. ucation and functional heterogeneity) and ego net- To ensure that the firms in the sample were sim- works (number of contacts, range of contacts, and ilar on basic environmental characteristics—espe- strength of ties) were drawn from TMT members’ cially a reliance on new knowledge—the sample and knowledge workers’ responses to question- firms conformed to Milkovich’s definition of high- naires. Measures of climate (for risk taking and technology firms as those “that emphasize inven- individualism) and knowledge creation capability tion and innovation in their business strategy, de- were drawn from the knowledge workers’ question- ploy a significant percentage of their financial naires.5 Organization-level scores were created for resources to R&D, employ a relatively high percent- experience, education, number of direct contacts, age of scientists and engineers in their workforce, strength of ties, and organizational climate by av- and compete in worldwide, short-life-cycle prod- eraging individual participant responses. We used uct markets‘ (1987: 80). average measures instead of additive measures be- Of the 211 technology firms contacted, 85 agreed cause the respondents were a representative sam- to participate in the study. Because of missing data ple of TMT members and knowledge workers. on some measures for 13 firms, the final sample Knowledge stocks. Stocks of organizational size was 72 companies (representing a 34 percent knowledge were measured in terms of the demo- participation rate). Organizations that agreed to graphic work experience (years in the industry) and participate did not differ from nonparticipants on formal education (years of post high school educa- either total revenue (t211  1.49, n.s.) or number of tion) of each of the TMT members and knowledge employees (t211  1.22, n.s.). Furthermore, on a workers who responded. We averaged years of in- limited sample for which data were available (84 of dustry experience and education for TMT members the 126 firms that did not participate), the level of and knowledge workers across respondents from 4 Across the companies in the study, TMT members 5 We conducted a principal axis factor analysis on the were corporate officers, including vice presidents of fi- aggregated organizational data to test for construct valid- nance, marketing, manufacturing, R&D, etc. knowledge ity and found support for a five-factor model (59.5% of workerss that were identified fell into the following cat- variance explained): stocks of knowledge, ego networks, egories: 23% were scientists or senior scientists, 43% climate for risk taking, climate for individualism, and were engineers, 11% were software developers, 10% knowledge creation capability. We found similar support were consultants, 8% were project managers, and 5% for a five-factor model using confirmatory factor analysis were marketing or sales personnel. TMT members and with individual data (analysis not shown: chi-square knowledge workerss were very similar in their back- goodness-of-fit test  434.55, p  .001, df  300, CFI  ground characteristics, except that the knowledge work- .87, GFI  .90, RMSEA  .06). These analyses provide ers had approximately two years less industry experience. evidence of convergent and discriminant validity. 352 Academy of Management Journal April the same firm to arrive at organizational scores. whereby network range is conceptualized as the Functional heterogeneity was measured with number of different status groups (categories of al- Blau’s (1977) heterogeneity index: (1  i 2), where ters) accessed by a network. As noted, 13 categories i is the proportion of the group in the ith category. of alters were identified, and we measured network A high score on this index indicates variability in range as the proportion of categories to which an functional backgrounds among respondents, or organization had at least one link. functional heterogeneity; a low score represents Strength of ties. As noted above, we asked re- greater functional homogeneity. spondents to identify the number of contacts that Ego networks. Measures of ego networks were they had for 13 categories of alters. In a second step collected from the surveys of TMT members and they identified the number of contacts within each knowledge workers. We used an ego-centered alter category that were critical to achieving the method that relied on a respondent—the ego—and company’s goals. Respondents were asked to list a set of alters who potentially had connections to the names of each of these important contacts and the ego (Wasserman & Faust, 1994). We identified then to answer a set of questions that focused on nine categories of external alters (representatives of these important relationships. Strength of ties was financial institutions, suppliers, customers, com- measured as an index that included the mean of the petitors, alliance partners, government agencies, duration of relationship, frequency of interaction, trade associations, boards of directors, and other) and emotional intensity of these key contacts. Du- and four categories of internal firm alters (people ration was measured as the average number of from operations, marketing/sales, R&D, and other). months that top managers’ and knowledge workers’ These categories were identified from theory, inter- relationships had existed. Frequency was mea- views with executives from high-technology firms sured as the average number of times per month the not in the study, and pretests with executive MBA managers and workers made contact with the al- students. ters. Emotional intensity was measured as the av- erage response, on a five-point scale, to the ques- Number of direct contacts. TMT members and tion: “How close is your relationship with these knowledge workers were asked about their rela- contacts on average.” The three items were stan- tionships with each of the 13 categories of alters. dardized and combined in a linear additive index More specifically, each respondent was asked to at the firm level of analysis. identify the number of direct contacts that he or she Organizational climate. Measures of climate had for each category of alter. The sum of contacts were collected from the surveys of knowledge across each of these categories was the total number workers. We used items from O’Reilly, Chatman, of contacts for each respondent. The number of and Caldwell’s (1991) instrument to measure the contacts was then measured as the average number climate dimensions of risk-taking and teamwork. of TMT and knowledge worker contacts across re- Climate for risk taking was measured with a five- spondents.6 Because the average number of con- item scale. Because the original teamwork factor tacts was not normally distributed, we used the comprised only two items, we developed a third natural logarithmic transformation of the average item to ensure more consistent measurement of the number of contacts in our statistical analyses. construct. Respondents were asked to assess their Network range. Since we were interested in an agreement (1  “strongly disagree,” to 5  organization’s ability to access diverse sets of “strongly agree”) with statements about their orga- knowledge, the network range measure was con- nization’s climate. Both scales showed good reli- structed to capture the breadth of knowledge that ability (risk taking,   .88; teamwork,   .87) and could be accessed through the aggregated different support for aggregation (see Bliese [1998]; risk tak- types of contacts. More specifically, we followed an ing ICC[1]  .22, ICC[2]  .73; teamwork ICC[1]  approach described by Burt and Minor (1983: 178) .25, ICC[2]  .71). Firm-level knowledge creation capability. Mea- sures of knowledge creation capability were pro- vided by the surveys of the knowledge workers. 6 The restriction of the number of contacts to a limited Drawing from the knowledge literature, we argued set of alters meant that we underestimated the total num- ber of contacts of individuals with large networks. How- that for effective exchange and combination to oc- ever, given our pretest and the wide range of alternate cur, individuals must: (1) have access to people or categories of alters, we felt that only less important con- groups with specialized information (access to par- tacts were likely to be omitted. Thus, we considered our ties); (2) be able to absorb and combine information approach a conservative test of the number of contacts. that has been exchanged (combination capability); 2005 Smith, Collins, and Clark 353 TABLE 1 Correlations, Means, and Standard Deviationsa Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 1. Knowledge creation capability 3.58 0.45 2. Years of industry experience 13.28 5.18 .16 3. Years of education 5.71 1.81 .37** .08 4. Functional heterogeneity 0.47 0.26 .02 .16 .13 5. Number of contacts 112.36 79.20 .26* .10 .20 .04 6. Network range 0.74 0.19 .18 .17 .23* .26* .31* 7. Network strength of ties 0.05 0.48 .34** .06 .16 .02 .14 .39** 8. Culture for risk-taking 3.20 0.72 .43** .06 .33** .07 .24 .11 .06 9. Culture for teamwork 3.53 0.70 .48** .10 .26* .11 .04 .08 .28* .44** 10. Number of employees 1,202.30 2,772.00 .25** .09 .34** .17 .14 .22 .19 .22 .18 11. R & D/sales 10.15 7.25 .20 .01 .12 .08 .17 .19 .09 .23* .10 .24* 12. Number of new products and 17.70 27.59 .47** .19 .29* .09 .22 .05 .31** .32** .27* .46** .35** services a n  72. * p  .05 ** p  .01 and (3) anticipate value from the exchange and control variables. We measured size using a natural combination process. We developed 15 five-point- logarithmic transformation of a firm’s number of scaled questions to measure the extent to which full-time employees. R&D spending was measured respondents had access, were capable, and antici- as a percentage of total sales. These measures were pated value from the exchange and combination obtained from company documents. process (contact the first author for the actual items). A first-order factor analysis with the 15 knowl- edge creation items showed that 12 items “loaded” RESULTS on a single factor, while three remaining items Table 1 reports the means, standard deviations, weakly cross-loaded on two additional factors (this and correlations of all variables. Table 2 reports the analysis is also available from the first author). results of regression analyses in which a firm’s When we removed the cross-loaded items, we knowledge creation capability and number of new found a single factor with an eigenvalue of 5.73. products and services are the dependent variables. Because there was strong evidence that these scales Overall, there is partial support for each of the might actually be a single factor, the 12 items were hypotheses. With regard to the stocks of knowledge combined into a single measure of knowledge cre- hypotheses (Hypotheses 1a–1c), we found that ation capability for each individual (  .87). In years of education was positively related to knowl- addition, there was strong support for aggregation (ICC[1]  .37, ICC[2]  .74). edge creation capability (  .34, p  .01); func- Number of new products and services. Our tional heterogeneity was marginally positively re- measure was the number of new products or ser- lated to it (  .18, p  .10); and experience was vices an organization had introduced in the most unrelated. recent year. In a meta-analysis of innovation stud- With regard to the ego network hypotheses (Hy- ies, Damanpour (1991) found that this count is a potheses 2a–2c), we found a positive relationship robust measure of innovation over a wide range of between the log of the number of direct contacts research settings. These data were collected during and firm knowledge creation capability (  .20, the interviews with the CEOs. This measure was p  .05). Consistently with Hypothesis 2c, the significantly correlated with the average percent- strength of network ties was positively related (  ages of sales spent on R&D (r  .35, p  .01), the .33, p  .01) to knowledge creation capability. Net- number of personnel assigned to R&D (r  .44, p  work range was not related to this capability. Both .01), and sales growth (r  .41, p  .01). of our measures of organizational climate were pos- Control variables. Organizational size and re- itively related to knowledge creation capability, search and development spending were used as supporting Hypotheses 3a and 3b (climate for risk 354 Academy of Management Journal April TABLE 2 Results of Regression Analysis Predicting Knowledge Creation Capability and Innovationsa Model 2: Number of Model 3: Number of Model 1: Knowledge New Products and New Products and Creation Capability Services Services Sobel Variables  t  t  t Test Number of employees .12 1.11 .38 4.67** .37 3.82** R&D/sales .08 0.77 .15 1.44 .16 1.57 Years of industry experienceb .04 0.36 .01 0.21 .06 0.61 0.35 Years of educationb .34 2.59** .21 1.97* .05 0.62 2.22* Functional heterogeneityb .18 1.93† .17 1.75 .10 1.05 1.76† Number of contactsc .20 2.22* .21 1.95* .03 0.42 1.97* Network rangeb .03 0.28 .03 0.24 .01 0.19 0.28 Network strength of tiesb .33 4.08** .26 2.48** .18 1.67 2.95** Culture for risk-takingd .37 4.62** .25 2.41** .15 1.44 3.14** Culture for teamworkd .32 3.92** .14 1.51 .04 0.35 2.89** Knowledge creation capabilityd .41 4.28** Adjusted R2 .51 .27 .36 F 18.44** 8.14** 12.03** a n  72. The dependent variable is the number of new innovations. b Data were provided by top management team members and core knowledge workers. Results based only on top management team data remained consistent. c Logarithm. d Data were provided by core knowledge workers. † p  .10 * p  .05 ** p  .01 taking,   .37, p  .01; climate for teamwork,   of new products and services. Following Baron and .32, p  .01).7 Kenny’s (1986) three-step procedure, we first exam- Table 2, model 3, also reports regression results ined the relationships between the independent with the number of new products and services as variables and the dependent variable. As shown in the dependent variable. Supporting Hypothesis 4, a Table 2, model 2, education, number of direct con- firm’s knowledge creation capability was positively tacts, strength of ties, and climate for risk taking related to its number of new products and services were significantly related to number of new prod- (  .41, p  .01). Other than organizational size ucts and services. Second, as demonstrated in Ta- (  .37, p  .01), none of the variables in the ble 2, model 1, a significant relationship exists be- model were directly related to the level of innovation. tween a firm’s stocks of knowledge, ego networks, We also expected that knowledge creation capa- and climate, and our mediator, knowledge creation bility would mediate the relationship between the capability. Third and finally, as model 3 in Table 2 independent variables (stocks of knowledge, ego demonstrates, we found that the previously signif- networks, and organizational climate) and number icant relationships between number of new prod- ucts and services and education, number of con- 7 Because measures of organizational climate and tacts, strength of ties, and climate for risk taking knowledge creation capability were provided from the were no longer significant when a firm’s knowledge same source, we performed an additional analysis creation capability was added to the equation. whereby we randomly selected half of the knowledge However, knowledge creation capability remained workers from each company as sources for climate mea- significantly related to number of new products sures and half of the sample as sources for the firm’s and services. The findings from this set of analyses knowledge creation capability. In models predicting suggest that knowledge creation capability fully knowledge creation capability, the beta for risk taking dropped from .37 to .35 (still significant), and the beta for mediates the relationship between years of educa- teamwork dropped from .32 to .30 (still significant). This tion, number of contacts, strength of ties, climate supplemental analysis suggested that common method for risk taking, and number of new products and variance did not account for our organizational climate services. knowledge creation results. Smaller samples, such as ours, make it difficult 2005 Smith, Collins, and Clark 355 to find direct relationships between independent ful in the invention of new knowledge. We concur variables and more distal dependent variables with Hansen (2002) that it would be interesting to (Shrout & Bolger, 2002), as the first step of the examine the nature of the knowledge that is trans- Baron and Kenny procedure outlined above dem- ferred through ego networks and explore the effect onstrated. Thus, we also tested the significance of that length of network connections has on knowl- the indirect effects of our independent variables on edge sharing. Moreover, one might usefully exam- a firm’s number of new products and services with ine the conditions under which the strength of ties a test designed by Sobel (1982; cf. Baron & Kenny, has negative impacts on knowledge creation, for 1986). The Sobel test is a more direct test of the example, when friendship or heightened familiar- mediation hypothesis because it examines the com- ity between actors gets in the way of their obtaining bined effects of the path between a dependent vari- useful information. In addition, since strong ties able and a moderator and the path between the are costly and may limit the ability of individuals moderator and an independent variable (Sobel, and organizations to build large networks, in future 1982). As shown in the final column of Table 3, we research, efforts should be made to identify which found that six of our independent variables had organizational ties should be strong and where or- significant or marginally significant indirect effects ganizations should instead leverage weak ties. We on number of new products and services through speculate that network range did not show signifi- knowledge creation capability. Results of the Sobel cant coefficients here because there was very little test were more robust than those of the Baron and variation in this measure at the firm level. Kenny (1986) procedure, suggesting that knowl- Our research also suggests that an organization’s edge creation capability mediated the relationship climate plays a strategic role in knowledge creation between years of education, functional heterogene- capability. We found that a climate that supports ity, number of direct contacts, strength of ties, cli- risk taking increases that capability. Perhaps em- mate for risk taking, climate for teamwork, and ployees of such firms are more open to new and number of new products and services. novel information and are more likely to interact in new ways, even when the payoff from such activi- ties is not certain. Similarly, firms that cultivate a DISCUSSION climate of teamwork are better able to stimulate This research was designed to answer two ques- exchange and combination between employees. tions: (1) How does the existing and accessible Thus, while individualism may have efficiency knowledge in a firm impact the firm’s knowledge benefits for organizations, teamwork and collective creation capability and (2) how does the knowledge action seem necessary for knowledge creation. creation capability affect the level of firm innova- Finally, we observed that existing and accessible tion? We found that certain aspects of existing and knowledge in a firm affects the rate of new products accessible knowledge did impact a firm’s knowl- and services entirely through the firm’s knowledge edge creation capability, which, in turn, impacted creation capability. This finding suggests that the level of new products and services introduced. knowledge creation capability may be necessary to Importantly, the research highlights the relation- successful innovation, and thus may be a key dy- ship between static or existing knowledge in a firm namic capability of firms. We treated existing and and the more dynamic knowledge creation capabil- accessible knowledge as static, measuring “stocks” ity. For example, we showed that stocks of em- reflecting current viewpoints and beliefs. On the ployee knowledge, measured as education level other hand, knowledge creation capability is dy- and functional heterogeneity, are related to the pro- namic, a process of combination and exchange cess of knowledge creation. From an organizational leading to new knowledge. Organizations’ attempts perspective, hiring and training well-educated em- to keep aligned with their environments may re- ployees with varying functional expertise seems to quire attention to both existing knowledge and the increase the likelihood that such employees will process of knowledge creation. combine and exchange their ideas to form new In conclusion, we examined the relationship be- knowledge. It would be interesting to explore if this tween a firm’s existing knowledge, its knowledge exchange and combination occur naturally when creation capability, and its new products and ser- knowledge stocks in an organization are high, or vice introductions. By empirically linking stocks of whether this process needs to be induced. knowledge, ego networks, and organizational cli- Our findings that number of direct contacts and mates to knowledge creation capability, we provide strength of ties were related to firm knowledge cre- a more complete picture of how firms create new ation capability support Nahapiet and Ghoshal’s knowledge and advance the knowledge creation (1998) contention that social networks can be use- literature. By connecting knowledge creation to 356 Academy of Management Journal April products/service introductions, we have demon- Brown, S., & Eisenhardt, K. 1995. Product development: strated that the study of the knowledge creation Past research, present findings, and future directions. capability has promise for increasing understand- Academy of Management Review, 20: 343-378. ing of how organizations evolve and adapt to their Burgelman, R. 1991. Intraorganizational ecology of strategy environments. Finally, by demonstrating that knowl- making and organizational adaptation: Theory and edge creation capability mediates the relationship be- field research. Organization Science, 2: 239-262. tween existing resources and product/service intro- Burt, R. S. 1982. Toward a structural theory of action. duction, we show how two somewhat divergent New York: Academic Press. streams in the knowledge literature, one somewhat Burt, R. S. 1992. Structural holes. Cambridge, MA: Har- static and the other dynamic, are related. However, vard University Press. more conceptual and empirical work is necessary if researchers are to fully understand how new knowl- Burt, R. S., & Minor, J. 1983. Applied network analysis: edge is created and exploited in organizations. A methodological introduction. Thousand Oaks, CA: Sage. Cohen, W. M., & Levinthal, D. A. 1990. Absorptive capacity: A new perspective in learning and innovation. Admin- REFERENCES istrative Science Quarterly, 17: 178-184. Amabile, T. M. 1988. A model of creativity and innova- Damanpour, F. 1991. Organizational innovation: A meta- tion in organizations. In B. M. Staw & L. Cummings analysis of effects of determinants and moderators. (Eds.), Research in organizational behavior, vol. Academy of Management Journal, 34: 555-590. 10: 123-167. Greenwich, CT: JAI Press. Dierickx, I., & Cool, K. 1989. Asset accumulation and Amabile, T. M. 1996. Creativity in context: Update to sustainability of competitive advantage. Manage- the social psychology of creativity. Boulder, CO: ment Science, 35: 554-571. Westview Press. Dougherty, D., Munir, K. & Subramaniam, M. 2002. Man- Argote, L. 1999. Organizational learning: Creating, re- aging technology flows in practice: A grounded the- taining and transferring knowledge. Boston: Klu- ory of sustainable innovation. Academy of Manage- wer Academic. ment Proceedings, Technology & Innovation Ashkanasy, N. M., Wilderom, C. P. M., & Peterson, M. F. Management Division: E1-E6. 2000. Introduction. In N. Ashkanasy, C. P. M. Wil- Drazin, R., & Rao, H. 2002. Harnessing managerial knowl- derom, & M. F. Peterson (Eds.), Handbook of orga- edge to implement product-line extensions: How do nizational culture & climate: 1-18. Thousand Oaks, mutual fund families allocate portfolio managers to CA: Sage. old and new funds? Academy of Management Jour- Bantel, K. A., & Jackson, S. E. 1989. Top management and nal, 45: 609-619. innovations in banking: Does the demography of the Ettlie, J., Bridges, W., & O’Keefe, R. 1984. Organization team make a difference? Strategic Management Journal, 10: 107-124. strategy and structural differences for radical versus incremental innovation. Management Science, 30: Baron, R., & Kenny, D. 1986. The moderator-mediator 682-695. variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Fiske, S., Kinder, D., & Larter, W. 1983. The novice and Journal of Personality and Social Psychology, 51: the expert: Knowledge based strategies in political 1173-1182. cognition. Journal of Experimental Social Psychol- ogy, 19: 381-400. Becker, G. S. 1964. Human capital. New York: Columbia University Press. Glaser, R. 1984. Education and thinking. American Psy- chologist, 39: 93-104. Blau, P. 1977. Inequality and heterogeneity: A primi- tive theory of social structure. New York: Free Granovetter, M. 1973. The strength of weak ties. Ameri- Press. can Journal of Sociology, 78: 1360-1380. Bliese, P. D. 1998. Group size, ICC values, and group- Hambrick, D., & Mason, P. 1984. Upper echelons: The level correlations: A simulation. Organizational organization as a reflection of its top managers. Research Methods, 1: 355-373. Academy of Management Review, 9: 192-206. Boeker, W. 1997. Executive migration and strategic Hansen, M. T. 1999. The search-transfer problem: The change: The effect of top manager movement on role of weak ties in sharing knowledge across orga- product market entry. Administrative Science nization subunits. Administrative Science Quar- Quarterly, 42: 231-236. terly, 44: 82-111. Boland, R. J., & Tenkasi, R.V. 1995. Perspectives making Hansen, M. T. 2002. Knowledge networks: Explaining and perspective taking in communities of knowing. effective knowledge sharing in multiunit companies. Organization Science, 6: 350-372. Organization Science, 13: 232-248. 2005 Smith, Collins, and Clark 357 Hargadon, A., & Fanelli, A. 2002. Action and possibility: H. Prosch (Eds.), Meaning: 22-45. Chicago: Univer- Reconciling dual perspectives of knowledge in orga- sity of Chicago Press. nizations. Organization Science, 13: 290-302. Schneider, B. 2000. The psychological life of organiza- Hargadon, A., & Sutton, R. 1997. Technology brokering tions. In N. Ashkanasy, C. P. M. Wilderom, & M. F. and innovation in a product development firm. Ad- Peterson (Eds.), Handbook of organizational cul- ministrative Science Quarterly, 42: 716-749. ture & climate: xvii-xxi. Thousand Oaks, CA: Sage. Huber, G. 1991. Organizational learning: The contribut- Shrout, P. E., & Bolger, N. 2002. Mediation in experimen- ing processes and the literatures. Organization Sci- tal and nonexperimental studies: New procedures ence, 2: 88-115 and recommendations. Psychological Methods, 7: Kimberly, J., & Evanisko, M. 1981. Organizational inno- 422-445. vation: The influence of individual, organizational Sobel, M. E. 1982. Asymptotic confidence intervals for and contextual factors on hospital adoption of tech- indirect effects in structural equation models. In S. nological and administrative innovations. Academy Leinhardt (Ed.), Sociological methodology: 290- of Management Journal, 24: 689-713. 312. San Francisco: Jossey-Bass. Kogut, B., & Zander, U. 1992. Knowledge of the firm, Starbuck, W. H. 1992. Learning by knowledge-intensive combination capabilities, and the replication of tech- firms. Journal of Management Studies, 29: 713-740. nology. Organization Science, 3: 383-397. Tushman, M. L. & O’Reilly, C. A. 1997. Winning through Kogut, B., & Zander, U. 1996. What firms do? Coordina- innovation: A practical guide to leading organiza- tion, identity and learning. Organization Science, 7: tional change and renewal. Boston: Harvard Busi- 502-518. ness School Publishing. Krackhardt, D. 1992. The strength of strong ties: The Wasserman, S., & Faust, K. 1994. Social network anal- importance of philos in organizations. In N. Nohria & ysis. Cambridge, England: Cambridge University R. G. Eccles (Eds.), Networks and organizations: Press. Structure, form, and action: 216-239. Cambridge, Weick, K., & Westley, F. 1996. Organizational learning: MA: Harvard University Press. Affirming an oxymoron. In S. R. Clegg, C. Hardy, & Levitt, B., & March, J. 1988. Organizational learning. In W. Nord (Eds.), Handbook of organizational de- W. R. Scott (Ed.), Annual review of sociology, vol. sign: 440-458. Thousand Oaks, CA: Sage. 14: 319-340. Palo Alto, CA: Annual Reviews. Lord, R., & Maher, R. 1990. Alternative information-pro- cessing models and their implications for theory, research and practice. Academy of Management Ken G. Smith (kgsmith@rhsmith.umd.edu) is the Dean’s Review, 15: 9-28. Chaired Professor of Business Strategy in the Robert H. Smith School of Business, University of Maryland. He McKeithan, K., Reitman, J., Rueter, H., & Hirtle, S. 1981. earned a Ph.D. from the University of Washington. His Knowledge organization and skill differences in research interests include competitive dynamics, dy- computer programmers. Cognitive Psychology, 13: namic capabilities, and strategic decision making. 307-325. Milkovich, G. T. 1987. Compensation systems in high Christopher J. Collins (cjc53@cornell.edu) is an assistant tech companies. In C. S. Anderson & A. Kleingartner professor of human resource studies in the College of (Eds.), Human resource management in high tech- Industrial and Labor Relations at Cornell University. He nology firms: 103-114. Lexington, MA: Lexington received his Ph.D. in organizational behavior and human Books. resources from the Robert H. Smith School of Business at the University of Maryland. His research interests in- Nahapiet, J., & Ghoshal, S. 1998. Social capital, intellec- clude strategic human resource management, firm inno- tual capital, and the organizational advantage. Acad- vation and knowledge creation, employee recruitment, emy of Management Review, 23: 242-266. and employment brand equity. Nemeth, C. 1992. Minority dissent as a stimulant to Kevin D. Clark (kevin.d.clark@villanova.edu) is an assis- group performance. In S. Worchel, W. Wood, & J. A. tant professor of strategic management in the College of Simpson (Eds.), Group process and productivity: Commerce and Finance at Villanova University. He re- 95-111. Newbury Park, CA: Sage. ceived his Ph.D. in strategic management from the Robert Nonaka, I., & Takeuchi, H. 1995. The knowledge creat- H. Smith School of Business at the University of Mary- ing company. New York: Oxford University Press. land. His research has investigated top management team O’Reilly, C. A., Chatman, J., & Caldwell, D. F. 1991. decision making in turbulent environments, the role of People and organizational culture: A profile compar- top management team social networks in performance, ison approach to assessing person-organization fit. and determinants of innovation and knowledge creation, Academy of Management Journal, 34: 487-516. particularly in technology firms. Polanyi, M. 1975. Personal knowledge. In M. Polanyi &